Joint Neural Architecture and Hyperparameter Search for Correlated Time Series Forecasting

This is the original pytorch implementation of SEARCH in the following paper: Joint Neural Architecture and Hyperparameter Search for Correlated Time Series Forecasting.

Requirements

• python 3.6
• see requirements.txt

Data Preparation

SEARCH is implemented on several public correlated time series forecasting datasets.

• PEMS03, PEMS04, PEMS07 and PEMS08 from STSGCN (AAAI-20). Download the data STSGCN_data.tar.gz with password: p72z and uncompress data file usingtar -zxvf data.tar.gz

• Solar-Energy and Electricity datasets from https://github.com/laiguokun/multivariate-time-series-data. Uncompress them and move them to the data folder.

Architecture Search

search for the best arch-hyper on the PEMS08 dataset

mkdir saved_model
CUDA_VISIBLE_DEVICES=0 python3.6 joint_search.py

Architecture test

test the arch-hyper searched on the PEMS08 dataset

cd test
CUDA_VISIBLE_DEVICES=0 python3.6 joint_test.py